How BYU Students Are Using Machine Learning to Transcribe Wilford Woodruff's Handwriting

Videographer: Connor Olvera

Speakers: Nathan Christiansen, Amber Oldroyd, Bryce Lunceford, Paul Smith

Technology Highlights

Transcript

Bryce Lunceford: We're working on something that hasn't, to my knowledge, hasn't really been done before.

Paul Smith: We're using machine learning to be able to read and translate Wilford Woodruff's words.

Nathan Christiansen: Basically we can take just a raw image of a document and get an initial transcription.

Bryce: So one of the things that we do is you get this loss. You kind of watch the network as it's training. And we have this thing called the loss function that basically tells you how bad is it at doing the task that you told it to do. And you watch as the loss goes down, which means it's getting less and less bad. So it's getting better.

Paul: We've had plenty of roadblocks. I mean, most recently we've realized that it's not a very good idea to train our model, to detect just lines first. This, it seems like it's doing better. Just going straight into what, you know, segmenting it into individual words.

Amber Oldroyd: This is an important problem for us to solve because we need to be able to grab all of the words from the letters in order to accurately represent the words of Wilford Woodruff and the words written to him.

Bryce: Something I've learned about Wilford Woodruff by doing this project is just how good he was. He, he exudes goodness.

Paul: I think it takes a bit of, a bit of vision, I think, right, to, to be able to sit back and think like, how can I help future generations understand what's going on? And to kind of think outside the box.

Amber: He took the time to write letters to a wide variety of people and just show his concern for them.

Bryce: He was always willing to help people. He talks about his experiences in his journals of going and giving blessings to the sick and just what an amazing man.